Dose or timing? Separating basal reduction from algorithmic suspension during fasting in type 1 diabetes: code, data and figures for a registered in-silico factorial study

Complete computational record for a registered within-subject factorial in-silico study of insulin dosing during fasting in type 1 diabetes, on the UVA/Padova model via simglucose 0.2.11.

This version (3.0.0) supersedes 1.0.0. It archives the full main campaign and the Amendment 2 addendum: 32 arms x 20 virtual patients x 7 fasting durations x 60 Monte Carlo replicates in the default cohort (268,800 simulations), 4 headline arms at 20 replicates in an illustrative regional cohort (11,200), and a 10,800-simulation registered addendum. 290,800 fourteen-day simulations in total, approximately 8,380 CPU-hours.

Contents. Run manifest with generation-time invariants; simulation harness; SLURM job-array scripts; merge and analysis code computing Monte Carlo standard error on paired contrasts at replicate level; the full 1,085-contrast table; the within-basal contrast family plotted in Figure 3; addendum analysis output including the H10 censored series; figure-generating code producing all five figures at 1200 PPI; and a pinned dependency lockfile.

Reproducibility. Seeds are a function of replicate and patient only and never of the arm, so every arm within a cell receives an identical noise realisation and all paired contrasts are formed on the same realisation. Both seed formulas are re-verified against the deposited manifest.

Licensing. Code under Apache-2.0; data, contrast tables and figure sources under CC-BY-4.0.

Not a medical device. Research and educational use only; not for clinical insulin dosing. simglucose is an open reimplementation of the UVA/Padova equations and is not the FDA-accepted T1DMS software.